Abstract: We make a few simple comparisons of the principles andperformance for noise reduction and edge detection withconventional methods versus neural network methods.Noise reduction methods discussed include the waveletpacket transform. Edge detection is discussed from thepoint of view of the Sobel and Canny transforms. Anapproach using the IBM ZISC036 neural network chip isalso discussed. In all cases, the results are compareto that of the biologically inspired PCNN. Anapplication of the `best if both worlds' isdemonstrated in a foveation/object isolationapplication for ATR.!8
展开▼